Chi-Square and PCA Based Feature Selection for Diabetes Detection with Ensemble Classifier

Diabetes mellitus is a metabolic disease that ranked among the top 10 causes of death by world health organization. During last few years, an alarming increase observed worldwide with 70% rise in since 2000 and 80% male deaths. If untreated, it results complications many vital organs human body which may lead to fatality. Early detection diabetes task significant importance start timely treatment. This study introduces methodology for classification diabetic normal people using ensemble machine learning model feature fusion Chi-square principal component analysis. An model, logistic tree classifier (LTC), proposed incorporates regression extra through soft voting mechanism. Experiments are also performed several well-known algorithms analyze their performance including regression, classifier, AdaBoost, Gaussian naive Bayes, decision tree, random forest, k nearest neighbor. In addition, experiments carried out analysis (PCA) (Chi-2) features influence selection on classifiers. Results indicate Chi-2 show high than both PCA original features. However, highest accuracy obtained when LTC used framework-work achieves 0.85 score available approaches prediction. statistical T-test proves significance approach over other approaches.